Hello Gabor!
I do not have much experience with distributed storage systems, but I think
that should work. The only problem might be the size of the individual AST
dumps.
We are going to check an idea for reducing a size: we can compress AST dumps while collecting them (it makes them nearly twice smaller) and uncompress them in runtime. Moreover, reading a smaller file and uncompressing it may be even faster than reading a big file from disk.